{"resourceId":"jisc-call-to-suppliers-genai-governance-cost-control-20261027","versions":[{"version":"external-97b46157e0e194d7734ef6ea364d1df1db3bdbc92e2487eb27ba6c281b67c72a","resource":{"id":"jisc-call-to-suppliers-genai-governance-cost-control-20261027","title":"Jisc and UK sector bodies set eight supplier expectations for generative AI: usage attribution, hard caps, no forced bundling, no double payment and model-change notice","organization":"Jisc (with Association of Colleges, GuildHE and Universities UK)","sector":"Higher and further education IT procurement","geography":"United Kingdom","publishedAt":"27 October 2026 (document dated as created 23 October 2026)","publicationDate":"2026-10-27","eventDate":"2026-10-27","sourceName":"Jisc repository","sourceLabel":"Sector statement (PDF)","sourceUrl":"https://repository.jisc.ac.uk/files/ai-licensing-statement.pdf","evidenceClass":"public-sector-association","outcomeClass":"emerging","topics":["governance-procurement","operating-model","infrastructure","data-security"],"finding":"Jisc, backed by the Association of Colleges, GuildHE and Universities UK, published expectations for AI suppliers covering commercial and governance terms. Suppliers are asked to avoid bundling AI in ways that obscure cost, preserve value through the term, give exportable and auditable usage data attributable to team, department, project and research group, and offer budget alerts, soft and hard caps, throttling and overage controls. It also asks for disclosure of which underlying models power embedded tools to avoid paying twice, advance notice and migration support for model retirement, and portability of prompts, agents and knowledge bases.","sledRelevance":"U.S. campus CIOs and consortia face the same metered-AI and embedded-AI cost questions; the eight areas translate into a negotiating checklist, with the limit that Jisc negotiates nationally and U.S. buyers mostly do not.","evidence":"Sector body statement of expectations; describes desired supplier behaviour, not observed results.","architectureImplications":"Usage telemetry must be exportable via APIs and attributable below tenant level; model routing and version continuity become contract terms.","governanceImplications":"Hard caps and delegated budget controls move AI spend governance into finance and unit leadership, not only IT.","securityPrivacyImplications":"Statement defers privacy and security to existing policies and supplier assurance; it adds data controls and portability on exit.","caveats":"Expectations only; suppliers have not committed. UK national buying model differs from U.S. state and consortium contracting.","streamIds":["campus-operations"],"roles":{"sales":"Sales framing: the UK sector has written down what buyers want from AI suppliers: cost attribution by department, hard caps, no forced bundling and no paying twice for the same model inside different products. Interpretation: U.S. CIOs, procurement officers and consortia can use the same list to audit their AI portfolio. A bounded offer is a contract and usage review against these eight areas. Do not claim savings; the statement contains no cost data or supplier commitments.","engineering":"Engineering framing: most expectations are technical. They require usage data exportable by API, attribution below tenant level, enable or disable controls per department or user, hard caps, and notice before model retirement. Interpretation: pre-sales should map which of a campus's AI tools can meet each item today and where embedded AI shares an underlying model. Treat it as a requirements list, not proof any vendor meets it.","delivery":"Delivery framing: turning the statement into practice means a cost-control operating model, with unit budgets, alert thresholds, monthly usage reports and a process for model changes. Interpretation: delivery teams should pilot attribution and caps with one or two heavy-use units before campus-wide chargeback. Acceptance should be defined as reports finance can reconcile and caps that actually stop spend, since the statement offers no benchmark."},"roleDetails":{"sales":{"customerProblem":"Unpredictable AI costs, bundled AI features and duplicate spend across tools built on the same models.","stakeholders":["Chief information officer","Procurement and contracts","Chief financial officer","Consortium buying groups","Research computing"],"discoveryQuestions":["Which of your current platforms have added AI features you pay for?","Can you attribute AI spend to departments today?","Do any contracts guarantee notice before a model is retired?"],"valueHypothesis":"Interpretation: a portfolio review against these expectations could expose duplicate or uncapped spend; no savings are shown in the source.","potentialEngagement":"A bounded AI contract and usage review scored against the eight Jisc expectation areas.","unsupportedClaims":["Specific savings","That suppliers will accept these terms","That UK national deals transfer to U.S. procurement"]},"engineering":{"fit":"Fits campuses running several AI tools and embedded AI features under separate contracts.","architectureIntegration":"Usage telemetry pulled by API into finance reporting, with tenant, department and user-level enablement and cap controls.","prerequisites":["Vendor admin APIs that expose usage","Identity groups mapped to cost centres","Finance data model for chargeback"],"constraints":["Many vendors expose usage only at tenant level","Embedded AI may hide which model it uses","UK framing; U.S. contracting differs by state"],"security":"Defers privacy and security to existing assurance; adds data export and portability on exit.","proofOfValue":["Test whether each tool's usage export can be attributed to a department","Confirm a hard cap stops consumption in a sandbox tenant"]},"delivery":{"work":"Stand up AI cost governance: attribution, caps, alerts and change notice handling.","dependencies":["Finance cost-centre mapping","Vendor contract amendments","Usage export access"],"ownership":"Central IT finance or IT business office, with unit budget holders owning their caps.","skillsAdoption":"Finance and IT staff need training on consumption billing and reading usage reports.","governanceCheckpoints":["Review of bundled AI features before renewal","Approval of unit caps","Model-change review before switching"],"acceptanceCriteria":["Proposed: monthly usage report reconciles to invoice within an agreed tolerance","Proposed: every AI tool mapped to an owner and cost centre"],"risks":["Vendors decline contract changes","Attribution data too coarse","Caps disrupt critical work if set too low"]}},"retrievedAt":"2026-10-27T21:50:45-05:00","enrichedAt":"2026-10-27T21:53:50-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Notes that gaps between consumer and enterprise offerings drive shadow usage, a staff-behaviour risk.","procurementImplications":"Proposes transparency at contract signature, protection from value erosion, and disclosure of underlying model supply chains to avoid double payment.","operatingModelImplications":"Treats AI as a consumption service that needs chargeback, forecasting and portfolio consolidation.","sourceVerification":{"openedUrl":"https://repository.jisc.ac.uk/files/ai-licensing-statement.pdf","referenceExcerpt":"Institutions need clear governance, predictable costs, meaningful oversight and confidence that investments will remain sustainable over time.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}